OpenAI Is Opening Up Advertising. Marketers Have Been Ready for Months

OpenAI’s road map has long resembled a whipsaw: fast product releases followed by faster cancellations (remember Custom GPTs?). That’s not terribly surprising. Sam Altman has promised fast iteration and, given his Y Combinator experience in coaching some of the hottest startups in tech, moving fast is to be expected. Still, for a product that reaches an estimated billion weekly active users, the changes come quickly.

This month, OpenAI finally released a self-serve advertising platform, a move that many analysts—myself included—have long expected. The role of AI in commerce is quickly evolving from product discovery to mid-funnel research and purchase influence. Whether the end state of agentic commerce stays a recommendation engine or evolves into fully automated robot shopping is still hotly contested.

Marketers Have Been Waiting

There’s no shortage of theories about AI’s potential in e-commerce, but the data is more interesting. Branch, a linking and measurement company, released results from a Q1 survey of enterprise executives on the role of AI in search in their businesses. The takeaway: marketers have been waiting for OpenAI to catch up.

First, marketers report AI-sourced traffic growing as a share of total traffic, reaching a mean of 35 percent last year. And their projections are bold: They expect 50 percent of their traffic to come from AI this year.

Not surprisingly, this is changing how dollars are spent. Sixty-five percent of respondents are dedicating at least a quarter of their marketing budget to AI; 28 percent are allocating over half. Quick back-of-the-envelope extrapolation against the global digital ad market ($791 billion) means that at a minimum, this represents $128 billion in investment this year.

The most striking finding is how fast respondents expect to move: Eighty-seven percent predict their company will execute closed-loop AI transactions this year. Incredible, especially given that the scaled technology to support this doesn’t exist.

The Challenge: Scale

Early reports indicated OpenAI’s platform is clunky, basic, and still showing signs of teething problems. That’s not surprising either. It takes years to build a successful, scaled ad model. Workflows, product development, and advertiser adoption aside, the very definition of a modern machine learning ad platform is built on data and learning from millions of user interactions. It took Meta three years to launch an ad product and another six to mature the model. And when Apple rewrote the targeting rules by removing deterministic identifiers, Meta spent two painful years rebuilding its ad targeting mechanisms.

But that’s not the biggest problem for advertisers. OpenAI has released a Conversions API (CAPI) and a cost-per-click (CPC) bidding model. Translation: They’re handing over advertising performance measurement to the advertisers. The burden of measuring success via CAPI isn’t new; Meta, Google, and all other modern ad platforms use one. The idea is that marketers send back alerts when conversions happen, which the advertiser uses to refine targeting and find better audiences. The CPC bidding model lets marketers tune campaigns to match their profitability metrics: the higher the bid, the more traffic.

The chief problem is that AI’s role isn’t yet defined relative to other channels. When a marketer records a conversion, they’ll send it to OpenAI, Meta, and Google. What happens if that user saw an ad on all three platforms? The advertiser has to decide how to distribute the shared credit—and in an emerging channel like AI-powered chatbots, how much credit should an LLM actually receive?

The same survey hinted at these issues. AI-sourced traffic was up 15 percent, but traditional SEO traffic was up 8 percent, reaching 53 percent as a share. This strongly suggests that AI isn’t replacing channels—it’s supplementing them. Meanwhile, 70 percent of marketers are struggling with the absolute basics of AI measurement. How do you measure AI traffic? Beyond a single UTM tracking flag, we have very little insight into how AI is influencing users. Today, most marketers are stuck asking the LLM where they rank, and then asking it how to do better.

The Future Is Promising

This isn’t a new problem. The advent of digital advertising gave us deterministic tracking, but the shift toward privacy—combined with the explosion of digital channels like web, search, social, mobile, CTV, and now AI—has eroded our ability to precisely track a unique purchase. This echoes the past, when Madison Avenue was pitching billboards, TV, and magazine ads. Marketers couldn’t track precisely across those mediums; they leaned on statistical models and incrementality testing to unearth the return on their advertising budgets. AI is emerging as a powerful, sticky, and valuable channel for marketers, but we’re going to need to learn how to measure its influence.

Adam Landis is the founder and CEO of AdLibertas, a mobile app data platform acquired by Branch in 2022. At Branch, he serves as head of strategic growth, where his deep experience in mobile advertising and data helps foster innovation in products that increase the ability to measure marketing performance in an increasingly difficult ecosystem.